Re: [PATCH 1/7] gdiplus: Make GpBitmap->bitmapbits access thread safe

[Dmitry Timoshkov]

Piotr Caban <piotr(a)codeweavers.com> wrote:
> +static inline BOOL image_lock(GpImage *image, BOOL *unlock)
> +{
> +    LONG id = GetCurrentThreadId(), v;
> +    v = InterlockedCompareExchange(&image->busy, id, 0);
> +    *unlock = !v;
> +    return !v || v==id;
> +}

Variable 'v' could have a better naming, for instance 'busy' or 'owner_tid'.
> +static inline void image_unlock(GpImage *image, BOOL unlock)
> +{
> +    if (unlock) image->busy = 0;
> +}

Is that even possible to get image_unlock() called with unlock != TRUE?

-- 
Dmitry.

[PATCH 1/7] gdiplus: Make GpBitmap->bitmapbits access thread safe

[Piotr Caban]

Hi Dmitry,


On 09/02/17 02:14, Dmitry Timoshkov wrote:
> > +static inline void image_unlock(GpImage *image, BOOL unlock)
> > +{
> > +    if (unlock) image->busy = 0;
> > +}
> > Is that even possible to get image_unlock() called with unlock != TRUE?

Currently not, but in theory  we should do something like this on all 
functions working on GpImage (I haven't tested all of them). For example 
it should be done in GdipGetImageWidth. This function is used internally 
so it will be useful to handle it. It also makes it harder to write 
incorrect code.

Thanks,
Piotr